Import AI 476: Creativity and RSI; worse-than-China open weight models; AI goes to space
New HSS benchmark shows best AI scores 53.6% vs human 93.1% on intuitive visual reasoning
“The two most common single mistakes were (1) missing a detail that decides the answer (21% of failures) and (2) misidentifying an object, person, or role (20%)”— Jack Clark
Elorian and Scale AI released Humanity's Sixth Sense (HSS), a 522-question benchmark testing intuitive visual reasoning across images and video; the best AI (GPT 6 Astra) scores 53.6% against a human baseline of 93.1%, with Chinese open-weight models (GLM-5.3-flash, Kimi K3) scoring below 27%. The benchmark is notable not just for exposing gaps but for showing frontier LLMs perform surprisingly well on tasks long assumed to require dedicated spatial world models, complicating the narrative that current architectures are fundamentally incapable of visual reasoning.